You are on page 1of 17

Board of Governors of the Federal Reserve System

International Finance Discussion Papers

Number 702

April 2001

updated Dec 2001


Francis E. Warnock

(forthcoming, Journal of International Money and Finance)

NOTE: International Finance Discussion Papers are preliminary materials circulated to stimulate
discussion and critical comment. References in publications to International Finance Discussion Papers
(other than an acknowledgment that the writer has had access to unpublished material) should be cleared
with the author or authors. Recent IFDPs are available on the Web at

Francis E. Warnock*

Abstract: The Tesar and Werner (1995) finding of very high turnover rates on foreign equity portfolios is
based on an underestimation of cross-border equity positions. Foreign turnover rates calculated using
information from comprehensive benchmark surveys on cross-border holdings are much lower than
previously reported and comparable to domestic turnover rates. However, the basic intuition from the Tesar-
Werner study, that transaction costs do not help explain the observed home bias, is confirmed using data on
transaction costs in 41 markets.

JEL Classification: G15, G11

Keywords: transaction costs, international portfolio diversification, turnover rates

The author is an economist in the International Finance Division of the Federal Reserve Board. I thank Eric
Boulay of Statistics Canada for help with Canadian IIP data and John Ammer, John Burger, Brian Doyle,
Caroline Freund, William Griever, Helene Rey, Linda Tesar, Charles Thomas, Veronica Warnock, and an
anonymous referee for helpful discussions and comments. All errors are my own. The views in this paper
are solely the responsibility of the authors and should not be interpreted as reflecting the views of the Board
of Governors of the Federal Reserve System, or of any other person associated with the Federal Reserve
System. email:
1. Introduction

Foreign equities comprise only a small portion of investors’ portfolios. For example, as

shown in Figure 1(a), foreign equities are now about 12 percent of U.S. investors’ equity portfolios,

a substantial increase from the one percent share two decades ago, but far smaller than the relative

size of non-U.S. equities in world market capitalization. Figure 1(b) condenses this information into

a measure of equity home bias, defined as one minus the ratio of the share of non-U.S. equities in

the U.S. and world portfolios. As the graph shows, the home bias in U.S. equity portfolios has

decreased substantially over the past two decades, but remains quite high.1

This paper focuses on another stylized fact of international finance, high turnover rates on

foreign equity portfolios, which is attributable to the evidence presented in Tesar and Werner (1995).

In particular, Tesar and Werner showed that in 1989 Canadians turned over their foreign equity

portfolio ten times faster than their domestic equity portfolio, and that U.S. residents turned over

their foreign portfolio at twice the rate of their domestic portfolio. High foreign turnover presented

a new puzzle for the theory of international portfolio choice. High turnover also ruled out high

transaction costs associated with trading foreign securities as a plausible explanation of home bias.

The Tesar-Werner findings on foreign turnover rates have generated two new bodies of

literature. First, models are now designed to produce high turnover on cross-border positions

(Rowland (1999), Coval (1999), and Guidolin (2001)). Second, the evidence against the plausibility

of transaction costs as a factor in home bias is cited as reason to dismiss transaction costs as a source

of barriers to international investment. This dismissal has led to models that rely on information

asymmetries to explain portfolio flows (Kang and Stulz (1997) and Brennan and Cao (1997)).

See Karolyi and Stulz (forthcoming) for a survey of the home bias literature.
The Tesar-Werner findings were based on data published before reliable cross-border

holdings data were available. Estimates of cross-border positions—the denominator in the turnover

rate on foreign holdings—were constructed from cumulated capital flows and estimated valuation

adjustments. Griever, Lee, and Warnock (2001) show that capital flows data are poorly designed

for estimating positions in foreign securities. This deficiency stems from the following sources. A

large component of the position is due to past valuation adjustments. Returns can vary substantially

across markets, which means the geography of the flows is a vital component of holdings estimates.

But this is precisely the shortcoming of the capital flows data. Because capital flows data were

designed according to the conventions of balance of payments accounting, they capture only the

country the transaction goes through, not the country of the issuer.

Comprehensive benchmark surveys of residents’ holdings of foreign equities, available for

a handful of countries, show the inaccuracies of past holdings estimates. The United States

conducted a benchmark survey in 1994. Based on the results of that survey, the Bureau of Economic

Analysis (BEA) increased their end-1993 estimate of U.S. holdings of foreign equities by $241

billion, or 80 percent. Such underestimations led to the Tesar-Werner result. Since the foreign

turnover rate is the value of transactions in foreign equities divided by the value of the foreign equity

position, understatement of the denominator generates an artificially high turnover rate. Once

estimates based on benchmark survey data are used, foreign turnover rates decrease substantially and

are roughly comparable to domestic turnover rates.

Section 2 documents this result for the United States and Canada for 1989, the year of the

Tesar-Werner data. Section 3 does the same for 1997, the year of the IMF-led Coordinated Portfolio

Investment Survey (CPIS), when both Canada and the United States conducted benchmark surveys.

Section 4 asks the following question. Is home bias a consequence of high transaction costs? Data

on actual transactions data for 41 countries suggests that the answer is no: Transaction costs are not

directly related to home bias. Section 5 concludes.

2. The Tesar-Werner Turnover Results Revisited

Tesar and Werner present three turnover measures. Domestic turnover is the ratio of annual

transactions on a market to its capitalization. The turnover rate in foreign equity held by domestic

residents is the ratio of annual transactions in foreign equities to the investment position in foreign

equities. Similarly, the turnover rate in domestic equity held by foreigners is the ratio of foreigners’

annual transactions in domestic equities to their holdings of domestic equities. This paper focuses

on the first two measures.

Table 1 shows turnover rates for 1989. Panels A and B give the original Tesar-Werner

turnover rates. Panel C shows that the foreign turnover rate for U.S. investors falls in half to 1.18,

and from 7.7 to 0.83 for Canadian investors, when more up-to-date estimates of cross-border

holdings are employed.2 In both cases, the sharp drop in the turnover rate was due to large upward

revisions in estimates of foreign equity holdings. For the United States, these holdings estimates

were more than doubled, from $92 billion, reported by Tesar and Werner, to $197 billion. For

Canada, the revised estimates are approximately ten times that reported in Tesar and Werner.

The point of this paper is not to fault Tesar and Werner or the international investment

position (IIP) data they used. The fact is, at least in the United States and likely elsewhere, capital

flows data are ill-suited to estimate positions in foreign equities. The geography is confounded, with

Our foreign turnover rates are comparable to those on Korean equities that are implied
by summary statistics presented in Choe, Kho, and Stulz (1999).

far too many transactions going through financial centers, making valuation adjustments—an

important component of holdings estimates—guesswork. Short of redesigning the portfolio flow

data to capture the foreign country in which the security was issued instead of the country through

which the trade was made, accurate estimates of foreign equity positions can only be obtained

through comprehensive, benchmark surveys.3

That said, the estimates of foreign holdings presented in the bottom panel of Table 1 are not

directly from benchmark surveys. The U.S. number is what the BEA now thinks—with the benefit

of information from the 1994 U.S. benchmark survey—U.S. holdings of foreign equities amounted

to in 1989. The survey gave a value as of March 1994; the end-1989 value, calculated by carrying

backward position estimates, is an estimate.4

The Canadian number is also an estimate, but a large part of it—Canadian holdings of U.S.

equities—is based on benchmark survey data. According to the 1989 U.S. benchmark survey of

foreigners’ holdings of U.S. securities, the market value of Canadian holdings of U.S. stocks at end-

1989 was $44 billion, or about C$51 billion. In addition, Canadian investors held other (non-U.S.)

foreign stocks that had an estimated book value of C$5 billion, or, based on a 2.87 price-to-book

ratio for non-U.S. securities, about C$14 billion in market value.5 My estimate of C$65 billion is

Total foreign holdings of domestic securities, but not the country-level detail, can be
more accurately estimated using capital flows data because the valuation adjustment does not
depend on correctly identifying the source country of the transaction.
See Bach (1997) for a description of the revisions to the U.S. IIP due to the 1994
benchmark survey.
Treasury Department (1998), a write-up of the 1994 Survey of Foreign Holdings of
U.S. Securities, also contains data from the 1989 U.S. survey. See Canada’s International
Investment Position (1995) for the 1989 Canadian data. The price-to-book ratio corresponds to
the MSCI (World ex US) Index.

the market value of Canadian holdings of U.S. equities (given by the U.S. benchmark survey and

published by Statistics Canada) plus the market value of Canadian holdings of non-U.S. foreign

equities (computed using the book value and price-to-book ratio).6

To restate, using information from two U.S. benchmark surveys of cross-border holdings—

the 1989 survey of foreign holdings of U.S. securities and the 1994 survey of U.S. holdings of

foreign securities—the 1989 turnover rates on Canadian and U.S. investors’ foreign equity portfolios

fall sharply from 7.7 and 2.5 to 0.8 and 1.2, respectively. In the next section, turnover rates for 1997,

when both the United States and Canada conducted benchmark surveys, are examined.

3. Turnover Rates Based on the 1997 CPIS

At the end of 1997, twenty-nine countries participated in the IMF-led Coordinated Portfolio

Investment Survey (CPIS), conducting simultaneous surveys to determine their residents’ holdings

of foreign securities.7 The CPIS should provide accurate measures of foreign holdings of investors

from twenty-nine countries and greatly enhance our knowledge of foreign turnover rates.

However, the data quality from the 1997 CPIS is not likely uniform. For many of these

countries, the CPIS marked a first attempt: Only one-third had previously reported an IIP statement.

Data collection approaches varied by country. The main choices countries had to make were whether

to (i) conduct the survey at the aggregate or security-by-security level, (ii) survey end-investors or

custodians, and (iii) make participation in the survey compulsory or mandatory. Surveying

Until 1997, Canadian IIP data for Canadian holdings of foreign stocks were reported
only at book value.
See IMF (2000) for a discussion of the coordinated surveys.

custodians (if domestic custodians exist), rather than just large end-investors, provides greater

coverage of households’ holdings (and retail holdings, in general), while a security-by-security

survey is likely to provide more reliable estimates than an aggregate survey. Countries that took the

aggregate approach asked the respondents to write down holdings by country. In contrast, in the

security-by-security approach, respondents provide security-by-security data on holdings. National

authorities then could cross-check the data to determine the accuracy of the value and country-

attribution of reported positions.

Most countries took an aggregate approach. Of those that conducted security-by-security

surveys, very few included data from custodians and obtained commercial databases to aid in their

cross-checks. Of those that did, to my knowledge only two, Canada and the United States, also

report data on gross transactions (i.e., gross purchases and gross sales) in foreign equities, which are

necessary to compute turnover rates. Restricting the analysis to those countries that followed best

practices and report gross transactions data rules out all countries but Canada and the United States.

Table 2 shows turnover rates on domestic and foreign equity portfolios for the United States

and Canada in 1997. Panel A shows that domestic turnover rates are low on the Toronto and New

York Stock Exchanges, but quite high on the Nasdaq. Panel B shows that while Canadians turned

over their foreign equity portfolio twice in 1997, this was due to a high turnover rate (2.76) on their

portfolio of U.S. equities; their turnover rate on non-U.S. foreign equities is about one.8 U.S.

investors turned over their foreign equity portfolio 1.3 times in 1997, comparable to their 1989

turnover rate. Thus, the table shows that investors may well turn over their foreign portfolios slightly

According to Statistics Canada, the reported market value of Canadian holdings of
foreign equities should be considered a lower bound, because the household sector is under-
represented in the survey.

faster than their domestic portfolios. It also highlights the fact that turnover rates vary greatly across

stock exchanges.

4. But Do Transaction Costs Matter?

Turnover rates on foreign equity portfolios are much lower than previously reported, but the

question remains: Can transaction costs explain the observed home bias in equity holdings?

Recently, researchers have investigated this question using a direct measure of transaction costs

faced by institutional investors across many countries. The measure, compiled for markets in 42

countries by Elkins-McSherry Co. and analyzed in Domowitz, Glen, and Madhaven (2000) and

Willoughby (1997), is comprised of three components: commissions, fees, and market impact costs.

Market impact costs, or liquidity costs, are intended to measure the deviation of the transaction price

from the price that would have prevailed had the trade not occurred. In practice, impact costs are

measured as the deviation of the transaction price from the day’s average price; see Willoughby

(1998) for a discussion.

Results in Domowitz et al. (2000) suggest that transaction costs cannot explain the home bias

in U.S. equity portfolios. Using cost-adjusted returns instead of unadjusted returns tilts the

composition of a U.S. investor’s global efficient portfolio from North America (which includes the

relatively high cost Nasdaq) toward Europe and Latin America, indicating that incorporating costs

makes the observed home bias even more of a puzzle.

Rather than working with cost-adjusted returns, Ahearne, Griever, and Warnock (2000) use

data from the 1997 benchmark survey of U.S. holdings of foreign equities—the same data used in

calculating the turnover estimates in Table 2—to investigate the relationship between transaction

costs and home bias. For 41 foreign countries, Figure 2 plots the Elkins-McSherry measure of

transaction costs for 1997 (normalized so that costs in the highest cost country, Korea, equals one)

against the country’s underweighting in U.S. investors’ portfolios, where underweighting (or bias)

is defined relative to the foreign country’s share of worldwide market capitalization. As the figure

shows, it is difficult to discern a simple bilateral relationship between trading costs and the measure

of bias; there is wide dispersion around a flat trendline.

While no direct evidence between transaction costs and home bias exists, there may well be

an indirect relationship. Since the NYSE is one of the lower cost exchanges in the world, one way

firms from high cost countries can alleviate trading costs in their stocks is by listing on the NYSE,

as in the model of Martin and Rey (2000).9 The general result from Ahearne et al. (2000) is that

countries whose firms tend to list on U.S. exchanges are less underweighted in U.S. portfolios. This

listing effect is greater for high transaction cost countries, suggesting that transaction costs may well

matter, albeit indirectly.

5. Conclusion

The Tesar-Werner home bias and high turnover puzzle is not evident when more up-to-date

and higher quality estimates of cross-border holdings are used. Turnover rates on foreign equity

portfolios are much lower than previously reported, and are roughly comparable to domestic turnover

rates. New data on transaction costs confirm the main Tesar-Werner conclusion that transaction

costs cannot explain the observed home bias.

See Alaganar and Bhar (2001) for evidence showing that Australian fund managers can
lower costs by using ADRs rather than the underlying Australian stock.

Perhaps more important than the findings is the message that estimates of cross-border

holdings can be inaccurate for the simple reason that capital flows data, designed based on the

conventions of balance of payments accounting, identify the country through which the transaction

was made. With inbound transactions data—that is, foreigners’ net purchases of domestic

securities—this is not a major obstacle for estimating aggregate positions. To estimate aggregate

foreign holdings of U.S. equities, for example, knowledge of the country of the foreign investor is

not necessary. We should be less confident, though, when estimating bilateral holdings, such as

German holdings of U.S. stocks. With outbound transactions data—that is, domestic residents’ net

purchases of foreign securities—the country of the issuer of the security is a vital piece of

information when estimating aggregate holdings of foreign securities. Since capital flows data do

not identify the country of the issuer, a price index to revalue holdings cannot be chosen with


Data quality should improve in the near future, because more countries are committing to

relatively frequent benchmark surveys of cross-border holdings using harmonized definitions.

Twenty-nine countries conducted outbound surveys at the end of 1997. Over 75 countries are on

board for an end-2001 survey.11 Thereafter, it is quite possible that annual surveys will be

conducted. Moreover, more countries will likely to be able to conduct a comprehensive, security-by-

security survey, which, according to IMF (2000), should provide more accurate results.

Canada is able to identify its residents’ transactions in U.S. securities. For all other
countries, though, Canada presents transactions data based on the country of the transactor.
See article in the IMF Survey (


Ahearne, A., W. Griever, and F. Warnock, 2000. Information costs and home bias: An analysis of
U.S. holdings of foreign equities. Federal Reserve Board, International Finance Discussion Paper

Alaganar, V., and R. Bhar, 2001. Diversification gains from American Depositary Receipts and
foreign equities: Evidence from Australian stocks. Journal of International Financial Markets,
Institutions and Money, 11: 97-113.

Bach, C., 1997. U.S. international transactions, revised estimates for 1974-96. Survey of Current
Business, 77(7): 43-55.

Brennan, M., and H. Cao, 1997. International portfolio investment flows. Journal of Finance,
52(5), 1851-1880.

Cho, H., B-C Kho, and R. Stulz, 1999. Do foreign investors destabilize stock markets? The Korean
experience in 1997. Journal of Financial Economics, 54(2), 227-264.

Coval, J., 1999. International capital flows when investors have local information. mimeo,
University of Michigan.

Domowitz, I., J. Glen, and A. Madhaven, 2000. Liquidity, volatility, and equity trading costs across
countries and over time. mimeo, Penn State University.

Griever, W., G. Lee, and F. Warnock, 2001. The U.S. system for measuring cross-border investment
in securities: A primer with a discussion of recent developments. Federal Reserve Bulletin, October.

Guidolin, M., 2001. Home bias and high turnover in an overlapping generations model with learning.
mimeo, University of Virginia.

International Monetary Fund, 2000. Analysis of the results of the 1997 Coordinated Portfolio
Investment Survey and plans for the next survey.

Kang, J., and R. Stulz, 1997. Why Is there a home bias? An analysis of foreign portfolio equity
ownership in Japan. Journal of Financial Economics, 46:3-28.

Karolyi, G.A., and R. Stulz, forthcoming. Are financial assets priced locally or globally? in
Constantinides, G., M. Harris, and R. Stulz, eds. Handbook of the Economics of Finance, North-

Martin, P., and H. Rey, 2000. Financial integration and asset returns. European Economic Review,
44: 1327-1350.

Rowland, P., 1999. Transaction costs and international portfolio diversification. Journal of
International Economics, 49: 145-170.

Tesar, L., and I. Werner, 1995. Home bias and high turnover. Journal of International Money and
Finance, 14: 467-493.

Treasury Department, 1998. Report on Foreign Portfolio Investment in the United States as of
December 31, 1997.

Treasury Department and Federal Reserve Board, 2000. United States Holdings of Foreign Long-
Term Securities as of December 31, 1997 and December 31, 1999.

Willoughby, J., 1997. Trade secrets. Institutional Investor. November, 69-75.

Willoughby, J., 1998. Executions song. Institutional Investor, 31(11), 51-56.

Table 1. Turnover rates in international equities, 1989 ($US billions unless otherwise noted)

A. Domestic turnover rates (from Tesar and Werner, 1995)

Total transactions on Equity Market Domestic Turnover
domestic market Capitalization (A/B)
(A) (B)
Canada 117.8 290.1 0.61
US 3223.9 3027.1 1.07

B. Turnover rates in foreign equity held by domestic residents (from Tesar and Werner, 1995)
Transactions in Investment positions Turnover rate
foreign equity in foreign equity (C/D)
(C) (D)
Canada 43.1 5.6 7.7
US 232.8 91.7 2.5

C. Turnover rates in foreign equity held by domestic residents (updated data)

Transactions in Investment positions Turnover rate
foreign equity in foreign equity (C/D)
(C) (D)
Canada (C$ billion) 54.3 65.4* 0.83
US 232.8 197.4 1.18

Estimates for Canadian holdings of foreign equities for 1989 are the author’s, based on data from the 1995 edition of
Canada’s International Investment Position. See text for complete discussion.

Table 2. Turnover rates in international equities, 1997 ($US billions unless otherwise noted)

A. Domestic turnover rates

Total transactions on Equity Market Domestic Turnover
domestic market Capitalization (A/B)
(A) (B)
Canada (Toronto) 305 568 0.54
US (NYSE) 5778 8880 0.65
US (Nasdaq) 4482 1726 2.60

B. Turnover rates in foreign equity held by domestic residents

Transactions in Investment positions Turnover rate
foreign equity in foreign equity (C/D)
(C) (D)
Canada (C$ billion)
all foreign equities 319 159 2.01
in US equities 227 82 2.76
in non-US equities 93 77 1.21
US 1553 1208 1.29

Notes and Sources: Panel A: Data are from the FIBV ( and are not directly comparable because Nasdaq
computes turnover rates differently from NYSE or TSE. The latter exchanges count as turnover only those transactions
which pass through their trading systems or which take place on the exchange's trading floor. Nasdaq includes in its
turnover figures all transactions subject to supervision by the market authority (transactions by member firms, and
sometimes non-members, with no distinction between on- and off-market and transactions made into foreign markets
reported on the national market). Transactions include trading in foreign firms listed on these exchanges and thus
overstate the turnover rates on domestic equities. Data for 1999 suggest that the degree of overstatement is quite small.
Panel B: Aggregate data for Canada are from Canada’s International Transactions in Securities and Canada’s
International Investment Position; both are Statistics Canada publications. The U.S./non-U.S. split for Canadian
holdings of and transactions in foreign equities is from unpublished data provided by Statistics Canada. U.S. data are
from and

Figure 1

Figure 1(a): Share of Foreign Equities in World and U.S. Portfolios




Share of Foreign Equities 50

in World Portfolio




Share of Foreign Equities 10

in U.S. Portfolio

1981 1985 1990 1995 2000
Sources: International Finance Corporation, International Federation of Stock Exchanges, and Federal Reserve Board.

Figure 1(b): Home Bias



Share of Foreign Equities in U.S. Portfolio

Home Bias = 1 - 0.6
Share of Foreign Equities in World Portfolio



1981 1985 1990 1995 2000

Figure 2: Transaction Costs and Home Bias

0.9 JP CL CZ
0.8 CH AT AU

0.7 LU





0 0.2 0.4 0.6 0.8 1

Relative Transaction Costs

Note: Bias, or underweighting in the U.S. portfolio, is one minus the relative weight of a
country’s equities in the U.S. portfolio to its weight in world market capitalization, as computed
in Ahearne, Griever, and Warnock (2000). The trendline from a regression of bias on transaction
costs is shown; the R2 of the regression is 0.00.

Country Codes
AR Argentina DK Denmark IN India PH Philippines
AT Austria EG Egypt IT Italy PK Pakistan
AU Australia ES Spain JP Japan PL Poland
BE Belgium FI Finland KR Korea PT Portugal
BR Brazil FR France LU Luxumbourg RU Russia
CA Canada GB Great Britain MA Morocco SE Sweden
CH Switzerland GR Greece MX Mexico SG Singapore
CL Chile HK Hong Kong MY Malaysia TH Thailand
CN China HU Hungary NL Netherlands TR Turkey
CO Colombia ID Indonesia NO Norway TW Taiwan
CZ Czech IE Ireland NZ New Zealand VE Venezuela
DE Germany IL Israel PE Peru ZA South Africa